Why it matters
Coding assistants often stumble on large, cross-file work; a repeatable spec-first workflow matters to teams trying to get reliable output from paid AI coding tools.
The tokenmaxxing angle
Spec-driven development front-loads planning to cut wasted generations and re-prompts, which is effectively a token-cost optimization: fewer failed AI attempts per feature shipped.
From the organizers
The listing says the talk argues AI coding assistants "excel at small tasks... but fall apart" on multi-file work, and teaches a process of writing a spec first, then having the AI build and iterate from it.